1. Field of the Invention
The present invention relates to an image-processing device, an image-capturing device, an image-processing method, and a recording medium that involve a restoration process based on a point spread function.
2. Description of the Related Art
In an object image to be taken through an image-capturing optical system, a so-called point spread phenomenon, in which a point object has a slight spread by the influence of the diffraction, aberration and others due to the image-capturing optical system, sometimes appears. A function indicating the response to a point light source of the optical system is referred to as a point spread function (PSF), and is known as a parameter that affects the resolution degradation (blur) of a taken image.
A taken image in which the image quality has been degraded because of the point spread phenomenon can recover the image quality by being subjected to a point-image restoration process based on the PSF. The point-image restoration process is a process in which the degradation property (point-image property) due to the aberration of a lens (optical system) and the like is determined in advance and the point spread of the taken image is cancelled by an image process using a restoration filter (recovery filter) corresponding to the point-image property.
For the point-image restoration process, various techniques have been proposed. For example, PTL 1 (Japanese Patent Application Laid-Open No. 2011-151627) discloses an image-processing device that can alter the power of the restoration process after the restoration process for an image is performed once. In the image-processing device, a correction filter is applied to a taken image, and thereby, a corrected image is generated. Then, a difference image between the taken image and the corrected image is generated, and the taken image, the corrected image and the difference image are saved.
Further, PTL 2 (Japanese Patent Application Laid-Open No. 2011-059813) discloses an image-processing device that performs an image recovery for an image after a non-linear correction, using a blind deconvolution. The image-processing device includes a correction section to perform a correction for reducing a non-linear gradation correction, for a taken image in which the non-linear gradation correction has been performed, and an image recovery section to perform the image recovery by applying the blind deconvolution to the taken image in which the gradation correction has been reduced.
Further, PTL 3 (Japanese Patent Application Laid-Open No. 2013-020610) discloses an image-processing device that reduces the excessive recovery of image data due to an image recovery process. In the image-processing device, an image recovery process is performed for an RGB-format color image data before a gamma process, the difference in the amplification and attenuation of a pixel signal value due to the gamma correction is absorbed, and the limiting value of the variation in the pixel signal value is calculated such that the maximum value of the variation is constant even after the gamma correction. This leads to the solution of technical problems such as “an occurrence of a situation in which because of saturated pixels, the degradation state of image data to be actually obtained does not coincide with the degradation state of image data to be intended as a recovery target of an image recovery filter”, and “an occurrence of an image quality degradation such as an undershoot or overshoot at an edge portion, particularly, the amplification of an undershoot at a low-luminance portion due to the gamma process after the image recovery process”.
Further, the point spread function for an optical system is used also in a restoration technology for an image in which the focal depth has been extended. For example, PTL 4 (Japanese Patent Application Laid-Open No. 2012-049759) discloses an image-capturing module that accurately executes an image restoration in a short time. In the image-capturing module, a restoration process is applied to a luminance signal after a demosaic (synchronization process). Therefore, it is unnecessary to have a parameter of the restoration process, for each of RGB, resulting in the speeding up of the restoration process. Further, adjacent pixels are arranged as a predetermined unit, a common restoration process parameter is applied to the unit, and a deconvolution process is performed. Thereby, the improvement of the restoration process accuracy is achieved.